diff --git a/litellm/router_strategy/complexity_router/complexity_router.py b/litellm/router_strategy/complexity_router/complexity_router.py
index d16063b9bd4..0573f8acf18 100644
--- a/litellm/router_strategy/complexity_router/complexity_router.py
+++ b/litellm/router_strategy/complexity_router/complexity_router.py
@@ -1020,13 +1020,14 @@ class ComplexityRouter(CustomLogger):
         weights: Final = self.config.dimension_weights
         weighted_score: Final = sum(d.score * weights.get(d.name, 0) for d in dimensions)
 
-        # Check for reasoning override (2+ reasoning markers)
+        boundaries: Final = self._effective_tier_boundaries()
+        scored_above_simple: Final = weighted_score >= boundaries["simple_medium"]
+
         # Reuse match count from _score_keyword_match to avoid scanning twice
-        if reasoning_match_count >= 2:
+        if reasoning_match_count >= 2 and scored_above_simple:
             return ComplexityTier.REASONING, weighted_score, tuple(signals), "reasoning_override"
 
         # Map score to tier
-        boundaries: Final = self._effective_tier_boundaries()
         if weighted_score < boundaries["simple_medium"]:
             tier = ComplexityTier.SIMPLE
         elif weighted_score < boundaries["medium_complex"]:
diff --git a/tests/test_litellm/router_strategy/test_complexity_router.py b/tests/test_litellm/router_strategy/test_complexity_router.py
index e1e8d9553b3..a148da8b675 100644
--- a/tests/test_litellm/router_strategy/test_complexity_router.py
+++ b/tests/test_litellm/router_strategy/test_complexity_router.py
@@ -267,6 +267,24 @@ class TestReasoningMarkerScoring:
         # 2+ reasoning markers should force REASONING tier
         assert tier == ComplexityTier.REASONING
 
+    def test_reasoning_override_does_not_rescue_a_simple_score(self, complexity_router):
+        """Reasoning markers on an otherwise trivial prompt must not reach REASONING."""
+        prompt = "hi, step by step, pros and cons"
+        tier, score, signals = complexity_router.classify(prompt)
+        assert score < complexity_router.config.tier_boundaries["simple_medium"]
+        assert any("step by step" in s and "pros and cons" in s for s in signals)
+        assert tier == ComplexityTier.SIMPLE
+
+    def test_reasoning_override_applies_at_the_simple_medium_boundary(self, complexity_router):
+        """A score sitting exactly on simple_medium is not SIMPLE, so the override still promotes it."""
+        prompt = (
+            "Give me the pros and cons, step by step, of moving our checkout service "
+            "to an event-driven architecture."
+        )
+        tier, score, signals = complexity_router.classify(prompt)
+        assert score == complexity_router.config.tier_boundaries["simple_medium"]
+        assert tier == ComplexityTier.REASONING
+
     def test_system_prompt_reasoning_not_counted(self, complexity_router):
         """Reasoning markers in system prompt should not count for override."""
         user_prompt = "What is 2+2?"
diff --git a/ui/litellm-dashboard/src/components/add_model/ClassificationMethodConfig.tsx b/ui/litellm-dashboard/src/components/add_model/ClassificationMethodConfig.tsx
index 8d42b34a14b..da0c50c7bd0 100644
--- a/ui/litellm-dashboard/src/components/add_model/ClassificationMethodConfig.tsx
+++ b/ui/litellm-dashboard/src/components/add_model/ClassificationMethodConfig.tsx
@@ -93,7 +93,7 @@ const HowClassificationWorks: React.FC<{ value: ComplexityRouterConfigValue }> =
           </li>
           <li>
             <strong>{effectiveTierLabel("REASONING", value.tier_labels)}</strong>: Score &gt; {ranges.complexReasoning}{" "}
-            (or 2+ reasoning markers)
+            (or 2+ reasoning markers with a score of at least {ranges.simpleMedium})
           </li>
         </ul>
       )}
diff --git a/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.test.tsx b/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.test.tsx
index a2afe0d7a0d..ed99e414706 100644
--- a/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.test.tsx
+++ b/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.test.tsx
@@ -59,7 +59,9 @@ describe("RoutingDecisionCard", () => {
         }}
       />,
     );
-    expect(screen.getByText("Heuristic, REASONING override (2 or more reasoning markers)")).toBeInTheDocument();
+    expect(
+      screen.getByText("Heuristic, REASONING override (2 or more reasoning markers, score above the lowest tier)"),
+    ).toBeInTheDocument();
     expect(screen.getByText("0.20")).toBeInTheDocument();
     // The score did not decide this tier, so NO band explanation may render at all.
     // Asserting the absence of one specific band would pass vacuously: 0.20 sits in
@@ -188,7 +190,9 @@ describe("RoutingDecisionCard", () => {
     // `signals` is gone under redaction; the cause alone must suppress the band.
     render(<RoutingDecisionCard decision={{ ...heuristic, cause: "reasoning_override", signals: undefined }} />);
     expect(screen.queryByText(/SIMPLE|MEDIUM|COMPLEX|at or above/)).not.toBeInTheDocument();
-    expect(screen.getByText("Heuristic, REASONING override (2 or more reasoning markers)")).toBeInTheDocument();
+    expect(
+      screen.getByText("Heuristic, REASONING override (2 or more reasoning markers, score above the lowest tier)"),
+    ).toBeInTheDocument();
   });
 
   it("shows the operator's tier name on the badge instead of the canonical one", () => {
@@ -212,7 +216,9 @@ describe("RoutingDecisionCard", () => {
     render(
       <RoutingDecisionCard decision={{ ...heuristic, cause: "reasoning_override", score: 0.2, tier_label: "Deep" }} />,
     );
-    expect(screen.getByText("Heuristic, Deep override (2 or more reasoning markers)")).toBeInTheDocument();
+    expect(
+      screen.getByText("Heuristic, Deep override (2 or more reasoning markers, score above the lowest tier)"),
+    ).toBeInTheDocument();
   });
 
   it("falls back to the raw cause for a value this build does not know", () => {
diff --git a/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.tsx b/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.tsx
index 0876a539653..cb680668849 100644
--- a/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.tsx
+++ b/ui/litellm-dashboard/src/components/view_logs/LogDetailsDrawer/RoutingDecisionCard.tsx
@@ -72,7 +72,7 @@ function describeCause(decision: RoutingDecision): string {
     case "heuristic_scorer":
       return "Heuristic scorer";
     case "reasoning_override":
-      return `Heuristic, ${tierLabel ?? "REASONING"} override (2 or more reasoning markers)`;
+      return `Heuristic, ${tierLabel ?? "REASONING"} override (2 or more reasoning markers, score above the lowest tier)`;
     case "llm_classifier":
       return classifierModel ? `LLM classifier (${classifierModel})` : "LLM classifier";
     case "literal_keyword_match":